feat(platform): add overnight v1 foundation

[TheophilusChinomona]

Switch discovery off Exa Websets to plain search so the local lead pipeline works on the current Exa plan. Add local Supabase migrations and provider plumbing needed for DeepSeek, Inngest, and safer retry/state handling.\n\nCapture the agreed product plan, current state, and overnight execution plan in root docs and docs/superpowers so the next session can resume from facts instead of scratch.

[vercel]

@TheophilusChinomona is attempting to deploy a commit to the CMN Labs Team on Vercel.

A member of the Team first needs to authorize it.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[greptile-apps]

Greptile Summary

The PR replaces Websets discovery with synchronous Exa search, adds shared OpenAI-compatible provider plumbing, improves enrichment retry state, and introduces local Supabase setup.

• Plain-search results are normalized and inserted directly as leads before enrichment events are published.
• AI calls now share configurable base URL and model selection for DeepSeek/OpenRouter compatibility.
• Local migrations establish the existing application schema, grants, and system-enabled control.
• Planning and current-state documents capture the proposed overnight implementation roadmap.

Confidence Score: 3/5

The PR should not merge until valid enrichment bios are preserved and failed enrichment-event publication becomes retryable.

The current parser deterministically discards generated lead bios, while transient Inngest publication failures leave newly created leads permanently pending even though discovery reports success.

Files Needing Attention: packages/autogtm-core/src/ai/enrichLead.ts; apps/autogtm/src/app/api/queries/_lib/startQueryRun.ts

Important Files Changed

Filename	Overview
apps/autogtm/src/app/api/queries/_lib/startQueryRun.ts	Implements synchronous discovery and direct lead creation, but silently loses enrichment work when event publication fails.
packages/autogtm-core/src/ai/enrichLead.ts	Ports enrichment to chat completions but drops the requested bio field during response parsing.
packages/autogtm-core/src/ai/client.ts	Centralizes OpenAI-compatible client and model configuration without an independently established defect.
apps/autogtm/src/inngest/functions.ts	Preserves enriched status when a completed lead is retried.
packages/autogtm-core/src/clients/exa.ts	Adds plain Exa search normalization for the existing lead-ingestion shape.
supabase/migrations/0001_autogtm_init.sql	Adds a local migration representing the application schema and existing broad authenticated-user RLS model.

_{Reviews (1): Last reviewed commit: "feat(platform): add overnight v1 foundat..." | Re-trigger Greptile}

[greptile-apps]

Enrichment parser drops bio

Every successful response is parsed through EnrichedLeadSchema, which omits the explicitly requested bio field and therefore strips it before persistence, leaving campaign personalization, digests, and the dashboard without the generated summary.

Suggested change

	const EnrichedLeadSchema = z.object({
	category: AllowedCategorySchema.catch('other'),
	full_name: z.string().catch('Unknown'),
	title: z.string().catch(''),
	bio: z.string().catch(''),
	expertise: z.array(z.string()).catch([]),
	social_links: z.record(z.unknown()).catch({}),
	total_audience: z.number().catch(0),
	content_types: z.array(z.string()).catch([]),
	promotion_fit_score: z.number().catch(5),
	promotion_fit_reason: z.string().catch(''),
	email: z.string().nullable().catch(null),
	});

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

[greptile-apps]

Dispatch failure leaves pending leads

When publishing the enrichment event batch fails, this block suppresses the error after the query has already been marked completed, causing the newly inserted leads to remain pending indefinitely without any retry or recovery path.